The goal of the proposed research is to understand the molecular basis of antigenic variation in the human influenza virus, and to apply this knowledge to improving influenza surveillance and vaccine strain selection. Antibodies directed against the surface protein hemagglutinin (HA) provide the major defense against influenza infection. Surveillance using genetic analysis of the quickly evolving HA provides a much finer scaled view of evolutionary change in circulating viral strains than traditional antigenic analysis. However, the relationship between genetic and antigenic variation is often unclear, and this relationship may vary over evolutionary time. Towards a better understanding of influenza evolution four specific aims are proposed: 1) A small number of codons in the HA of influenza A subtype H3N2 appear to have been under positive selection by the human immune system to repeatedly change the amino acid they encode. A model for predicting evolution has been developed which assumes a selective advantage for newly emerging viruses with additional changes at these codons. This model performed successfully in retrospective tests. In Aim 1 viruses predicted to be "most fit" at the end of each influenza season are tested to determine whether they caused the next epidemic. 2) This prediction model assumes that codons identified as having been positively selected in the past will continue to be positively selected in the future. This assumption is tested in Aim 2 using datasets drawn from different influenza seasons. 3) Hypotheses about the types and amounts of genetic change needed to produce an antigenically distinct HA will be developed through observations of past epidemics. Reverse genetics will be used to test these hypotheses: the HAs of transfectant viruses will be mutated at selected codons and then contrasted for differences in antigenicity. Genetic change that alters antigenicity will be characterized using homology models of the HA. One potential outcome will be the incorporation of homology modeling into current techniques of influenza surveillance. 4) The generality of the results from study of H3N2 will be tested by contrasting them with similar studies of influenza A subtype H1N1 and influenza B.